Dialectic Group / Perpetual Fund / August 2026 / Confidential Research

Thesis & Valuation
Memorandum

Probabilistic sum-of-the-parts valuation Q3 2026 Post-Q2 Earnings Update

$3.93-$4.32 TEquity Value Range
$286-$314/shareValue Per Share Range
$71.7 BFY2026E Economic Revenue
Space
Connectivity
AI & Digital
Page 02

Executive Summary

Two months after the largest IPO in history, the market has handed the opportunity back to buyers. Our week-1 prediction held true to form ($166-258/share), but it is increasingly clear that market observers have difficulty valuing this complex conglomerate of the future. We posit that near-term SpaceX is an AI infrastructure firm with a satellite connectivity business that is supported by launch services. While these 3 primary revenue streams: terrestrial datacenters, Starlink, and Launch services areas are projected to be the bulk of the $71.7B in 2026 economic revenue, we note with great interest the acceleration of Grok & Cursor post-acquisition.

SPCX closed today at $134.13 (~$1.80T market capitalization) below its IPO print and beneath the $166-$258 entry range we framed in June. Even as the first public quarter validated nearly every operating assumption in our model, revenue is up 92% year-over-year, the first quarter of material company-wide Adjusted EBITDA ($3.5B, +191%), Starlink subscribers grew to 12.0M, and most importantly nameplate compute up 3.5x to 1.4 GW, $14.1B of Cloud Services Agreements signed in ninety days, and $100B of balance-sheet liquidity following the IPO and the inaugural $25B investment-grade bond. Our post-Q2 announcement recalibration now drives every major segment off its relevant physical unit of production, including tonnes to orbit, subscribers, gigawatts of installed compute, all materially lifts our projections. We are considerably more bullish on this asset than we were at the IPO. Blended fair value rises to $286/share (blended equity $3.93T), the Monte Carlo P50 to $314/share ($4.32T equity), with right-skew to $629/share ($8.64T) at P90.

Our action here is to continue to accumulate SPCX aggressively at current levels while employing our advanced yielding strategies on top of the asset. Our long-term valuation target of $10T ($748/share) is unchanged, but is deepened in its substantiation by near-term business lines rather than exciting long-term plans such as Terafab, Orbital data centres and deep space optionality. All the myriad possibilities of space are effectively a low-cost option at these levels.

We also open the discussion on the best path forward for a merger between SpaceX and Tesla, which present important synergies along with a brief note on potential other synergistic transactions such as The Boring Co. over time. We’d wager a guess that the combined entity ends up being referred to just as ‘X’ and likely trades under said ticker.

Grok Bot is just the latest in accelerating network effects across the business lines currently operated under the SpaceXAI brand. SpaceX is pulling mankind to the future, more than any other company on the planet or off planet and we consider exposure to this asset fundamental for our future-forward investment portfolio.

Key Model Outputs (Post-Q2 2026 Model)

Blended Fair Value: $286/share | Blended Enterprise Value: ~$3.867T | Blended Equity: ~$3.927T (horizon weights 65 / 30 / 5)

Monte Carlo P10 / P50 / P90 equity: $1.504T / $4.319T / $8.641T (5,000 trials)

Market reference: $134.13/share (~$1.80T) | 2026E economic revenue $71.7B | 2026E EBITDA $46.3B | Long-term target $10T ($748/share)

Q2 2026: The First Public Quarter

Q2 ended June 30, 2026 - Revenues of $7.8B grew 92% year-over-year and 66% sequentially. Adjusted EBITDA of $3.5B nearly tripled. The AI segment swung from $(609)M of Adjusted EBITDA in Q1 to +$1.1B in Q2 was the fastest segment-level EBITDA inflection we have seen at this scale, driven by Cloud Services Agreements converting compute capacity into contracted revenue almost as fast as it is energized. Connectivity grew revenue 66% with income from operations up 79%, and Space completed two Starship V3 test flights while sustaining a ~156-launch annualized cadence with 1,041 tonnes delivered to orbit in the first half.

Three disclosures changed our model more than any headline number. First, SpaceX now reports mass to orbit (485t in Q2) and launch counts by customer versus internal, allowing us to rebuild Launch Services off tonnage rather than launch-price abstractions. Second, the AI segment discloses nameplate compute in gigawatts (1.4 GW, up from 0.4 GW a year ago) allows us to drive Terrestrial Data Centre TDCs such as Colossus and Macrohard off installed GW at a market-clearing revenue per gigawatt (~$13-14B/GW/yr, the rate implied by the $27.8B/yr of contracted tenancy against ~2 GW exiting 2026). Third, subscriber and ARPU disclosure (12.0M subs, $66 flat) confirms Starlink is running the launch playbook in telecom: hold price flat against collapsing unit costs, and let volume compound. Every major segment in our model is now driven off its referent physical unit of production with group revenues totalling $800B in 2030.

The uncommon read on the quarter is in what GAAP conceals. Twenty-eight of the thirty-eight Q2 launches carried internal payloads and produced almost no reported launch revenue, of which GAAP shows $648M of launch services revenue while 397 tonnes of internal deployment was, in economic substance, capital formation for Connectivity and AI. Mass to orbit is therefore the single best leading indicator of future segment revenue, and it is reported a full year before it monetizes. Similarly, the quarter's $15.8B of AI capex was matched almost one-for-one by $14.1B of freshly contracted CSAs, a book-to-build ratio near parity that converts what bears call capex risk into what landlords call pre-leasing. SpaceX has quietly built a second toll road beside access to orbit, including access to tokens: rival AI labs including Anthropic, Google, Reflection now rent xAI's landlord infrastructure just as rival constellations ride SpaceX rockets. The company collects the toll either way. We anecdotally note that observers of the Sport of Kings (Foundation Models) seem to note that companies prefer SpaceX data-centres over peers, which could imply a long-term premium to market for data-centre rental. Currently we are seeing that play out.

The balance sheet ended the quarter with $100.0B of cash and marketable securities against roughly $39.5B of debt and finance leases, with a net cash position of $60.5B, with $47.5B of backlog. The inaugural $25B bond priced at a 5.855% weighted-average coupon across tranches out to 2056: credit markets are pricing SpaceX as investment-grade infrastructure while the equity markets price it below our bear-adjusted fair value. Interesting guidance given by the company that they believe they can capture a “significant percentage of global connectivity market”, expectation of “more than a (1) launch per day in 2027” and an expectation to “bring 5-10GW of compute online by YE27 and a total 20GW of compute online by YE2028” represent incredible upside for this opportunity.

Page 03

Investment Thesis

For a deep dive on the pillars of our investment thesis, please review the pre-IPO analysis here. SpaceX is rapidly accelerating its strategic advantages across the different business units in the major overarching business segments of space exploration, connectivity, and AI. We further partition nine separate businesses across these categories. The Q2 print upgraded each pillar from thesis to evidence.

Page 04

Business Unit Analysis

SpaceX is a complex conglomerate operating, by some analysis, more than 2 dozen different businesses across the 3 major categories (Space, Connectivity, AI). We separated these major categories into nine discernible business units and worked through revenue and profitability potential on 3 different horizons blending 3 different outcome scenarios (bear, base, bull), while heavily weighting near-term potential for the sake of conservatism. We have rebuilt each segment's projection off its underlying physical metric, launches and tonnes to orbit, subscribers and ARPU, gigawatts of installed compute, rather than abstract revenue growth rates and arrived at a blended probability-weighted revenue in 2030 of approximately $800B which appropriately balances SpaceX’s guidance of $1T in revenue. It is Elon’s job to paint an inspiring future and lay out lofty goals as it serves as a rallying cry that pushes his people forward aggressively. It is our job as analysts to pair these goals to physical constraints and appropriately discount BHAGs (big hairy audacious goals) given Elon’s track record of delivering the impossible, somewhat late.

Page 05

Valuation Analysis

Methodology

We primarily employed a bottoms-up, probability-weighted sum-of-the-parts (SOTP) DCF across nine business units, each modeled with three scenarios (Bear / Base / Bull) and a materialization probability, with strong overweighting of near-term cash flows. Post-Q2, every major segment build is driven off its physical unit of production.

Comparables Analysis

We treat trading comparables as a cross-check and a downside bound, not as the primary basis for value. SpaceX has no clean peer and it is so obviously positioned to be the largest most important company on the planet that we find ourselves in full-throated belly laughs when talking with traders who short this asset. A blended whole-company multiple is therefore directional at best, which is why our central value rests on the sum-of-the-parts DCF.

Viewed across the relevant peer sets, the comparables span an extremely wide band. Defense primes trade at roughly 2x revenue and anchor low; high-growth space names (Rocket Lab, AST SpaceMobile) anchor high; the AI cohort sits between OpenAI ~28x revenue, Anthropic ~21x, Palantir ~47x, NVIDIA ~22x. The arithmetic has now inverted in the buyer's favour: at $134.13 the market values SpaceX at ~$1.80T of roughly 25x our 2026E economic revenue of $71.7B and only ~11x our 2027E of $169.5B, against an AI-lab median of ~24.5x. On reported GAAP (which excludes internal-payload economics), the annualized Q2 run-rate of ~$31B implies ~57x and the spread between the two multiples is itself the mispricing: the market is paying GAAP multiples for a company whose economic output is more than twice its reported revenue.

The synthesis is that comparables price the company as it is today, not the convergence flywheel we underwrite, and they penalize SpaceX for being mid-cycle in the heaviest investment phase in corporate history ($28.5B of H1 capex against $47.5B of backlog and $100B of liquidity). They remain most useful as a soft floor near $0.9-1.0T, a level the market briefly rediscovered and from which the Q2 print should force a re-rating.

Elon Premium

The Elon Premium has vanished and with it part of the upside opportunity. In June, the pre-IPO marks implied ~1.8x the AI-lab median multiple; at $134.13, SpaceX trades at ~21x our 2026E economic revenue, in line with the OpenAI/Anthropic median (~24.5x) and roughly half Palantir's ~47x, despite growing 92% with a launch monopoly, the world's fastest-scaling telecom, and a tenant-financed compute build attached. On our updated numbers the market is charging nothing for Musk, nothing for the flywheel, and nothing for eight of the nine business units: the Connectivity segment alone supports the current enterprise value at our SOTP base case ($3.06T of the $4.49T total).

The Elon Premium is best understood as compensation for the compounding optionality of an operator who moves faster, builds more ambitiously and competently, and controls more of the value chain than any single peer. SpaceX has no clean comparable and for the first time since the 2024 tender rounds, the market is not asking to be paid for that fact.

CAPEX

Given the exceptional profitability of TDCs which accentuate SpaceX’s engineering and execution advantage and the hard lessons of hardware won through many years of hard execution we would expect CAPEX spend to nearly double versus the guided numbers over the balance of this decade. We would expect that 1 trillion in CAPEX is spent by YE2030 and note that such a build will likely require dilutive financing during this period along with a significantly higher debt load. This will add short term pressure to valuation but the dividend in the end is quite literally out of this world.

Official Actuals — CapEx by Filing Segment ($B)
Filing segment Basis 2025 FY Q1 2026 Q2 2026 H1 2026
SpaceFiled3.831.051.172.23
ConnectivityFiled4.181.331.372.70
AIFiled12.737.7215.8323.55
TotalFiled20.7410.1118.3728.48

H1 already 83% AI. Musk: high AI spend for at least two more quarters. FY26 run-rate ~$55–70B.

Goldman Sachs case — Total Planned CapEx ($B)
Business Filing 2026 2027 2028 2029 2030 5-year
Launch ServicesSpace4.56.06.05.05.026.5
StarlinkConnectivity5.58.012.014.016.055.5
StarshieldConnectivity1.01.52.02.02.08.5
xAIAI1.01.52.02.02.08.5
Terrestrial Data CentresAI48.0121.0101.065.050.0385.0
Orbital Data CentresAI0.52.08.012.015.037.5
X PlatformAI0.30.30.30.30.31.5
TerafabAI2.010.015.015.013.055.0
Deep Space OptionalitySpace0.00.00.01.02.03.0
Total planned capex62.8150.3146.3116.3105.3581.0

$360B cumulative through 2028, ~80% AI.

SemiAnalysis / Musk GW case — Total Planned CapEx ($B)
Business Filing 2026 2027 2028 2029 2030 5-year
Launch ServicesSpace4.56.06.05.05.026.5
StarlinkConnectivity5.58.012.014.016.055.5
StarshieldConnectivity1.01.52.02.02.08.5
xAIAI1.01.52.02.02.08.5
Terrestrial Data CentresAI55.0350.0150.0100.080.0735.0
Orbital Data CentresAI0.52.08.012.015.037.5
X PlatformAI0.30.30.30.30.31.5
TerafabAI2.010.015.015.013.055.0
Deep Space OptionalitySpace0.00.00.01.02.03.0
Total planned capex69.8379.3195.3151.3135.3931.0

$50B per GW, 6–8 GW incremental in 2027.

Dialectic bull case — Total Planned CapEx ($B)
Business Filing 2026 2027 2028 2029 2030 5-year
Launch ServicesSpace4.58.09.08.08.037.5
StarlinkConnectivity5.510.014.020.026.075.5
StarshieldConnectivity1.02.02.53.03.512.0
xAIAI1.01.52.02.02.08.5
Terrestrial Data CentresAI55.0350.0150.0100.080.0735.0
Orbital Data CentresAI0.52.08.015.022.047.5
X PlatformAI0.30.30.30.30.31.5
TerafabAI2.010.020.025.035.092.0
Deep Space OptionalitySpace0.00.00.01.02.03.0
Total planned capex69.8383.8205.8174.3178.81,012.5

Five-year total $1,012.5B. 2030 is $178.8B so the year row foots to the exhibit total.

Page 06

The SpaceX–Tesla Merger: Strategic Positioning

A merger with Tesla seems obvious at this point, pending regulatory approval. The strategic reasons in favour of the merger are listed in order of importance;

  1. Elon’s control and focus: we discuss Elon’s relative ownership and voting control in greater depth below. But it is clear that Elon’s ownership of Tesla is a sensitive point that could be partially rectified by a merger at the right relative valuation. Also, having most of his assets under one roof will allow for the best focused output, something the world has still not seen.
  2. If AI + Robotics will dominate the world, the physical synergies across Connectivity (SpaceX) + Robotics (Tesla) + AI (SpaceX + Tesla) accelerated by Terafab (joint venture) are very compelling. The new Tesla Roadster that apparently will have cold propulsion technology from SpaceX collab is the first example of literally dozens of engineering synergies between the companies that will take us into the future.
  3. Current crossover between the companies: Terafab is designed around AI5-class FSD silicon and AI6-class Optimus inference chips; Colossus trains the models that Tesla's fleet and robots will run at the edge; Starlink D2C provides the connectivity layer for a global robotics fleet; that means Optimus and the FSD automotive fleet; and both companies draw on the same energy, manufacturing, and talent stack. We have excluded the merger from our valuation and every dollar of it is upside to the numbers in this memo. We’ve seen many instances over the years where the SpaceX engineering team is used by Elon like a SEAL Team-6 to go in and solve intractable challenges in unimaginably short periods of time. The consolidation and proliferation of this corporate culture of ‘a shocking zone of competence’ should not be understated.

Voting control across relative valuations

Elon holds 83.5% of SpaceX voting power through super-voting Class B shares against 19.9% of Tesla's single-class votes. We take as axiomatic and believe the board does too and that Musk should not emerge from any combination with less than 56-58% of voting control: enough to be unassailable through the AGI build-out and the coming decade of capital raises, Terafab's $119B full buildout, ongoing compensation awards), with buffer above a bare majority. The table below computes Musk's post-merger voting power across relative valuations of the two companies under the two candidate structures: SpaceX acquiring Tesla in an all-stock exchange (new SPCX Class A issued to Tesla holders; Class B super-voting intact), versus any single-class combination, which includes every structure in which Tesla is the acquirer.

Musk post-merger voting power by relative valuation and deal structure
Relative valuation (SpaceX : Tesla) Implied Tesla equity value New SPCX Class A issued (B sh) Tesla holders’ % of combined Musk voting — SpaceX acquires (dual-class) Musk voting — single-class / Tesla-led
75 : 25$600B4.525%78.0%36.5%
70 : 30$771B5.730%76.6%35.4%
65 : 35$969B7.235%75.0%34.3%
61 : 39 (market)$1,151B8.639%73.7%33.4%
55 : 45$1,473B11.045%71.5%32.1%
50 : 50$1,800B13.450%69.4%31.0%
45 : 55$2,200B16.455%67.1%29.8%
40 : 60$2,700B20.160%64.5%28.7%

Under a SpaceX-led dual-class exchange, Musk lands between 64.5% and 78.0% across every plausible ratio, the 56-58% floor is cleared everywhere and would bind only if Tesla commanded roughly 70% of the combined entity which seems categorically impossible given the execution and value creation of SpaceX. However, we do note that the Roadster demo, Robotaxi roll-out or Optimus robots beginning sales could inflect TSLA upwards and there may be volatility in the ratio. Under any single-class structure, Musk lands between 28.7% and 36.5% at every ratio: the floor is not merely missed, it is unreachable at any price.

Prediction markets agree on the direction and disagree only on the clock: Polymarket currently prices a 5% probability of consummation by September 30, 2026 and 18% by year-end with a curve whose mass sits in 2027, consistent with the natural sequencing of lockup expiry, initial index inclusion, and a re-rated SPCX currency.

We would forecast that a merger is likely in the second half of 2027.

Drawbacks to a potential merger are that the market is unlikely to understand this incredibly complex mega-conglomerate and investors are unlikely to see a premium on the combined assets for some time. Also that the regulator ire will also be more focused and likely increased, especially if there is a change in US leadership. And Musk’s remuneration will likely need to be reworked and may have a dilutive impact for shareholders (acceptable).

An interesting thought experiment emerges from the merger analysis and if long term investors are optimizing for Elon’s rightful control (and thus focus, etc.), doesn’t it make sense to roll in The Boring Co. and/or Neuralink over time? Elon holds approximately 80% of The Boring Co. and an estimated 50% of Neuralink, which would make both transactions accretive for him and the enterprising investor that holds exposure to both assets, which we do and are very high conviction on. We estimate that a split across the 4 organizations of 80% SPCX, 15% TSLA, 3% Boring and 2% Neuralink could be a prudent approach to gain accretive upside in a full roll-up which pairing for relative risks and risk-adjusted potential of each related company.

Page 07

Risks & Key Considerations

Page 08

Conclusion

As Peter Diamandis aptly puts it, it is the most exciting time in history to be alive and this is the most exciting organization in history. The initial Q2 report served to substantiate our pre-IPO analysis: 92% revenue growth, a company-wide EBITDA inflection, doubling subscribers, tripling compute, and $100.0B of gross liquidity set against an external financing requirement of roughly $238.3B through 2029. We reiterate a blended fair value of $286/share, a Monte Carlo P50 of $314/share. We find it interesting that the market isn’t pricing the convergence of SpaceX’s advantages against its peers.

At today’s levels one is acquiring a leading AI compute business in TDCs plus a profitable launch business and connectivity network, all of the inspiring possibilities that the moon and mars bring; in addition to the only fully vertically integrated AI stack and robotics (with Tesla) are included for free.

We reiterate and upsize our accumulation plan and yielding plan accordingly asserting that the fair valuation can be achieved over 2-3 years and maintain a projected 10T+ enterprise valuation by the early 2030s.

Disclaimer: Dialectic is an investor in SpaceX, Tesla and related entities discussed herein. This document is not financial advice or a recommendation or solicitation for investment and the financial modelling herein may represent projections or biases based on our own interpretation of current and potential events using public sources and may be incorrect. Dialectic makes no representations of any kind. This is purely for informational purposes.

Page 09 / Annexes A-G

Data Room